Basin-Related Effects on Ground Motion for Earthquake Scenarios In

Geophys. J. Int. (2006) 166, 197–212 doi: 10.1111/j.1365-246X.2006.02909.x

Basin-related effects on ground motion for earthquake scenarios in the Lower Rhine Embayment ∗ Michael Ewald,1 Heiner Igel,1 Klaus-GünterHinzen2 and Frank Scherbaum3 1Department of Earth and Environmental Studies, Section Geophysics, Ludwig-Maximilians-University München, Germany 2Abt. fürErdbebengeologie, University Cologne, Germany 3Institut fürGeowissenschaften, University of Potsdam, Germany

Accepted 2006 January 11. Received 2006 January 8; in original form 2005 March 20 Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

SUMMARY The deterministic calculation of earthquake scenarios using complete waveform modelling plays an increasingly important role in estimating shaking hazard in seismically active regions. Here we apply 3-D numerical modelling of seismic wave propagation to M 6+ earthquake scenarios in the area of the Lower Rhine Embayment, one of the seismically most active regions in central Europe. Using a 3-D basin model derived from geology, borehole information and seismic experiments, we aim at demonstrating the strong dependence of ground shaking on hypocentre location and basin structure. The simulations are carried out up to frequencies of ca. 1 Hz. As expected, the basin structure leads to strong lateral variations in peak ground motion, ampliﬁcation and shaking duration. Depending on source–basin–receiver geometry, the effects correlate with basin depth and the slope of the basin ﬂanks; yet, the basin also affects peak ground motion and estimated shaking hazard thereof outside the basin. Comparison with measured seismograms for one of the earthquakes shows that some of the main characteristics of the wave motion are reproduced. Cumulating the derived seismic intensities from the three modelled earthquake scenarios leads to a predominantly basin correlated intensity distribution for our study area. Key words: deterministic shaking hazard, earthquake scenarios, numerical wave propagation, seismic intensities, site effects. GJI Seismology

also been studied for various earthquakes in the vicinity of the Los INTRODUCTION Angeles Basin (e.g. Vidale & Helmberger 1988; Scrivner & The study area, the Lower Rhine Embayment (LRE) is part of the Helmberger 1994; Olsen et al. 1997; Olsen 2000). European rift system (Fig. 1) and shows moderate intraplate seis- The LRE was subject to intensive studies on ground motion am- micity. On average, several earthquakes of magnitude 5 and higher plification, especially after the damaging 1992 Roermond, M 5.9 occur every century (Reamer & Hinzen 2004). The LRE cuts as earthquake. Most of these investigations focused on 1-D estimation a sedimentary basin into the Rhenish Massif with a general NW– of site amplification based on spectral ratios of local seismicity and SE strike of the major faults. In its western part, the Roer Val- ambient noise measurements (e.g. Ibs-von Seht & Wohlenberg 1999; ley Graben (RVG) sediment layers exceed 1.5 km thickness. The Scherbaum et al. 2003; Ohrnberger et al. 2004; Hinzen et al. 2004; whole basin is densely populated and highly industrialized with Parolai et al. 2005). While these techniques give valid estimates on large petrochemical and chemical plants. Basins of such structure are 1-D amplification factors, they do not provide information on 3-D known to significantly amplify ground motions during earthquakes. effects, such as wave front focusing due to low-velocity zones or Striking examples are the damages in the Mexico City, M 8.1, edge-diffracted waves. 3-D wave propagation simulations account Michoacan earthquake (Sing & Ordaz 1993; Cárdenas& Chávez- for such issues and can, therefore, provide additional information on Garcia 2003) and in parts of San Francisco during the 1989 Loma basin-related seismic hazard. The 1995 Hyogo-Ken Nanbu, M 6.9 Prieta, M 7 earthquake (Stidham et al. 1999). The variability of earthquake near the city of Kobe, Japan, drastically proved the im- ground motion due to lateral variations in sediment thickness has portance of 3-D effects on the wavefield (e.g. Kawase 1996; Pitarka & Irikura 1996a). Numerous other studies concentrated on this issue (e.g. Boore 1972; Olsen et al. 1995a,b; Pitarka & Irikura 1996b; Graves 1998; Satoh et al. 2001). ∗ Corresponding author: Department fürGeo- und Umweltwissenschaften, Several recent studies on southern California and the Los Sektion Geophysik, Theresienstr. 41 80333 München,Germany. E-mail: Angeles basin showed significant lateral variations of ground mo- [email protected] tion due to 3-D geologic structures (e.g. Olsen et al. 1997; Olsen

C 2006 The Authors 197 Journal compilation C 2006 RAS 198 M. Ewald et al.

2000; Peyrat et al. 2001; Eisner & Clayton 2002; Komatitsch et al. as the LRE and a moderate seismicity leading to earthquakes with 2004). Incorporating complex kinematic finite sources, advantage predominantly dip-slip mechanisms. Arrows in Fig. 1 indicate the is taken of the detailed knowledge of the corresponding earthquake main directions of extensional stress in this region. Stress inversion source processes due to dense strong motion observations as well as from fault plane solutions of 110 earthquakes in the northern Rhine the high-resolution 3-D model of the Los Angeles Basin (Magistrale area (Hinzen 2003) indicates a subvertical direction of the largest et al. 1998, 2000; Süss& Shaw 2003). principal stress in the LRE. However, in the neighboring Stavelot- In this pilot study of earthquake ground motion in parts of the Venn Massif and Rhenish Massif the maximum principal stress is LRE the basin structure is represented by a 3-D model with depth subhorizontal. The maximum horizontal compressional stress in the dependent velocity and uniform attenuation upon layered bedrock. area has a trend of 125◦. The scenario earthquakes are modelled as finite-source moment tensors. We present three simulations of damaging earthquakes that oc- Seismicity curred in the LRE within the last 250 yr using a 3-D staggered-grid Together with the Belgian zone, the RVG belongs to the seismically finite-difference technique (e.g. Igel et al. 1995; Graves 1996). The most active onshore zones in northwestern Europe. Its active tec- Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021 spatial and temporal discretization allows us to simulate the wave- tonic extension regime makes it to one of the few areas in Europe field for frequencies up to 1 Hz. The main goal of this study is to north of the Alps producing damaging earthquakes. From histori- investigate the complexity of the resulting wavefield and the effects cal records more than 20 damaging earthquakes in the past 300 yr of amplification and prolonged shaking due to 3-D effects of the are known (Schwarzbach 1951). Local magnitudes of the strongest sedimentary basin in the study area. The key questions we want to historic earthquakes, 1976 Dürenand 1692 Verviers were estimated address are: (1) How does the basin model alter the peak values of to 6.4 and 6.8 (Hinzen & Oemisch 2001), respectively. Probabilis- ground motion? (2) Where—with respect to the basin geometry— tic seismic hazard analysis of this region estimates an intensity VII are amplifications to be expected? (3) How does the basin affect occurrence rate of 10−3 yr−1 (Rosenhauer & Ahorner 1994). The hazard estimates derived from peak ground motion? main present day seismic activity is observed in the RVG and its sur- The results from this study reveal strong lateral heterogeneity roundings. Within the last century 38 earthquakes with magnitudes in the distribution of peak ground velocities (PGV) and shaking M L ≥ 4.0 occurred in the northern Rhine area (Reamer & Hinzen duration due to the basin topography and the source location. Seis- 2004). Since 1755, the RVG and its extensions have experienced mogram sections illustrate the different effects of amplified and at least nine earthquakes with M L ≥ 5.0. The earthquakes of 1756 prolonged shaking and demonstrate the intrinsic 3-D nature of the Düren, M L 6.1, 1951 Euskirchen, M L 5.7 and the 1992 Roermond phenomena. Regions dominated by intraplate tectonics are charac- event M L 5.9 have been chosen as scenarios in this study. The M L terized by low deformation rates and long recurrence times for large 4.9 Alsdorf, 2002 July 22, event is by now the only larger local earthquakes. Consequently, the strong motion database is sparse for earthquake recorded by the seismic network in the LRE. Analysis such areas and shaking hazard assessment is afflicted with large un- of this earthquake and modelling of the observations allowed the certainties. Our results, therefore, further emphasize the need for appraisal of the current basin model (Ewald 2005). It was shown detailed 3-D basin models and accurate ground motion modelling that—despite of local discrepancies—peak ground motions for fre- in order to assess seismic hazard in the LRE and similar regions of quencies up to 1 Hz were reproduced for stations located throughout interest. the basin with a maximum deviation factor of two.

STUDY AREA—THE LOWER RHINE Basin model EMBAYMENT The sediments in the Lower Rhine Embayment have accumulated Location since the Carboniferous age. This long period results in a complex stratigraphy with locally varying thickness of the different sequences The LRE is a sedimentary basin located in the northwestern part of (Geluk et al. 1994). In this study the stratigraphy is represented by Germany, parts of the Netherlands and Belgium. Its shape is roughly a 3-D structure of soft sediments embedded by a layered bedrock a northwest opening triangle. The northern extension is called West model. Seismic velocities within the sediments increase smoothly Netherlands Basin ending at the North Sea coast. The close-up map with depth. As velocity depth function we used the relation in Fig. 1 illustrates its location in Central Europe, while the map in = ∗ + 0.37 . Fig. 2 gives an impression of the dense urban agglomeration in this VS 118 (z 1) (Parolai et al 2002) area. for the shear waves,√ P-wave velocities were derived assuming a V P /V S ratio of 3. For computational reasons the model was limited to a lowest shear wave velocity of 1000 m s−1. Although recent Tectonics research supports lower shear wave velocities in the Cologne Basin The RVG is located within the northern parts of the Cenozoic rift (Hinzen et al. 2004) this model was chosen as a compromise, in system of central and Western Europe, which spreads over 1100 km order to fulfill computational requirements as well as account for from the North Sea coast to the Mediterranean. A detailed discus- the gradient character of sediment velocities with depth. Attenua- sion about the Cenozoic rift system can be found in Ziegler (1994). tion within the sedimentary layer is modelled with a uniform quality Seismicity in this RVG is driven mainly by regional plate interactions factor of Q S = 50 for shear waves and Q P = 100 for P waves, using leading to a maximum horizontal compressional stress in NW–SE a standard linear solid approach (Robertsson et al. 1994). direction. As the RVG is roughly parallel to this direction, it shows A depth-to-basin map was compiled from seismic profiles as a significant active extension (Geluk et al. 1994). The result of this well as borehole data with horizontal resolution of about 1500 m extension regime is the accumulation of thick sedimentary basins (Scherbaum et al. 2003; Hinzen et al. 2004). The model parameters

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS Basin effects on ground motion in the LRE 199 Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

Figure 1. Geographic location of the study area. The Lower Rhine Embayment developed inside the Lower Rhine Graben, the northwestern part of the Rhine Graben System. Direction of regional maximum compressive crustal stress is NW–SE, resulting in a zone of tension oriented as indicated by black arrows (Hinzen 2003). The main tectonic blocks are divided by normal faults. The location of several large earthquakes within the study area is shown (asterisks).

are outlined in Table 1. Fig. 2 includes a contour plot of the basin The Roermond earthquake of 1992 April 13, with highest inten- model. sities I MM = VII and local magnitude M L = 5.9 is located near the town of Roermond in the southern part of the province Limburg in the Netherlands. It ranges among the strongest instrumentally Simulated earthquake scenarios observed earthquakes in Central Europe. The estimated recurrence interval for an earthquake of this size in the LRE is ca. 200 yr In order to demonstrate combined path and site effects and the re- (de Crook 1994). Source parameters for the Roermond event are esti- sulting complexity in basin amplification the chosen events are all mated in various studies (e.g. Ahorner 1994; Braunmiller et al. 1994; located close to the basin edges but in different epicentral regions. Pelzing 1994). Scherbaum (1994) discussed the intrinsic variability In the following the main features of these events are summarized: of the source parameters using a stochastic method. The earthquake The Düren earthquakes of 1755 December 27 and 1756 scenarios were modelled as circular finite-sources with source radii February 18, with highest intensities estimated from historical derived with Brune’s model using an assumed stress drop of 10 bars records as I = VIII, local magnitudes M = 5.8 and M = MM L L (Brune 1970). The source parameters of the modelled earthquakes 6.1, respectively, occurred 10 km south-west of the town of Düren are summarized in Table 2. between the LRE and the mountain range of the Stavelot-Venn Mas- sif. Their epicentres are close to each other and are most probably correlated with the active Feldbiss fault system, one of the south- SIMULATION RESULTS western border faults of the RVG (Meidow & Ahorner 1994). The earthquake of 1756, which is chosen as one scenario in this study, is Visco-elastic simulations of the earthquake scenarios described the strongest historical event in this region. The Euskirchen earth- above are employed using a finite-difference scheme on staggered quake of 1951 March 14, reaching intensities of I MM = VII–VIII, grids (e.g. Igel et al. 1995; Graves 1996). The model space was and local magnitude M L = 5.7 is located at the western border of discretized into ca. 417 million grid cells with a uniform horizontal the LRE near the town of Euskirchen. spacing of 100 m and a varying vertical spacing ranging from 50 m

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS 200 M. Ewald et al. Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

Figure 2. Detailed map of the study area. Red stars indicate epicentres of the simulated earthquakes. Greyscale shading depicts regional topography whereas urban areas are coloured in dark grey. The sedimentary basin depth is indicated by dashed isolines in 100 m intervals. The basin reaches a maximum depth of about 1900 m. Seismic network stations are marked as stars and by their station codes. A synthetic proﬁle A–A with 30 receivers is shown as black dots.

Table 1. Model parameters. Table 2. Earthquake Source Parameters. Layer Depth P-wave velocity S-wave velocity Density D¨uren Euskirchen Roermond [km] [m s−1][ms−1] [kg m−3] Date 1756 February 18 1951 March 14, 1992 April 13, Sediment 0–1.9 1730–3110 1000–1800 2200 Latitude 50◦ 45 50◦ 38 51◦10 Bedrock 0–1 5500 3175 2800 Longitude 6◦ 21 6◦ 44 5◦ 56 1–2 5750 3320 2800 Depth [km] 14 9 17 2–4 5980 3452 2800 Strike [‘] 135 110 120 4–10 6120 3533 2800 Dip [‘] 70 80 70 10–18 6300 3637 2800 Rake [‘] 270 270 260 18–26 6360 3672 2800 Rise Time [s] 1.0 1.0 1.0 16 26–30 7350 4244 2800 M 0 [10 N m] 14.0 3.7 7.5 Source Radius [m] 3900 2500 3200

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS Basin effects on ground motion in the LRE 201

Table 3. Summary of simulation parameters. the steep flank of the basin (E). Profiles for the north–south com- Spatial discretization [m] horizontal 100 ponent (Fig. 5) show similar effects as for the east–west horizontal Spatial discretization [m] vertical 50... 120 component. For the Dürenearthquake this component reveals very Temporal discretization [s] 0.0198 low frequency content in the northern part of the basin due to su- Lowest S-wave velocity [m s−1] 1000 perposition of trapped waves that was notable in the snapshots. In Grid size (physical model) 1400 × 1400 × 180 Fig. 6 the same profile sections are shown for the vertical ground Grid size (computational model) 1444 × 1444 × 200 motion. This component significantly reveals the dependence of am- Model size (physical) 140 km × 140 km × 30 km plitudes on basin depth. Especially for the Euskirchen simulation Number of time steps 16960 strong resonant amplification is present. Simulation time [s] 60 Earthquake scenario simulations can be calibrated and verified Memory usage [GB] 110 by comparison with observations. Unfortunately, most of the data Computation time [h] Using 32 12 nodes of the Hitachi SR8000 at recorded for the 1992 Roermond earthquake are clipped for stations Leibniz Rechenzentrum, München located inside the study area. Even though some similarities between observations and simulations are shown below, these should not be

overemphasized. Fig. 7 compares the three component synthetics Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021 in the sedimentary layer to 120 m in bedrock. With the numer- with observations for the station TGA at an epicentral distance of ical scheme used waveform modelling was possible for frequen- 56 km near the city of Bergheim (Fig. 2). Synthetics and observa- cies up to 1 Hz. A summary of modelling parameters is given in tions show similarities in four important features: (1) matching P- Table 3. and S-wave onsets in all components; (2) amplitudes are reproduced 3-D simulations of wave propagation provide the complete wave- with a maximum difference of ca. 50 per cent; (3) the amplitude ratio field over the whole model area. This allows gathering different between P and S waves is similar. For both horizontal components aspects of ground motion, such as peak values, static deformation the synthetic ratio (peak-to-peak amplitudes used) is within 20 per and/or rotation, at any point in time and space. In this section, recent of the observed values and (4) envelopes that are influenced by sults from simulations of the scenario earthquakes are presented the source mechanism and the underground structure look similar with special emphasis on the lateral variation of the effects at the except for the EW component. An excellent match in waveforms surface as a function of source location with respect to the basin is found for the P- and S-wave arrivals on the north–south com- structure. ponent and also late arriving energy on the vertical component is In Fig. 3 snapshots of the EW velocity component at the sur- reproduced. face for the three earthquakes are shown. Basin-related effects on Numerical simulations allow the mapping of ground motion pa- surface ground motion are particularly visible by comparing waves rameters such as PGV.From these PGV maps, seismic intensity can propagating into and away from the basin. Specific details of the be estimated and compared with observed intensities. For historical wavefield are marked with labels (A–D). Especially for the Eu- earthquakes macroseismic intensities are usually the only source of skirchen event with its epicentre south of the basin the resulting information. Seismic intensities can be estimated from PGV through wavefield splits into a nearly undisturbed part south of the basin empirical relations according to the following equations: and the scattered wavefield within the basin north of the epicen- = . + . tre. Diffraction and scattering of seismic waves inside the basin is IMM 2 10 log (PGV) 3 40 for intensities below V,and = . + . prominent for the Dürenearthquake (A). Trapped energy and rever- IMM 3 47 log (PGV) 2 35 for intensity V . . berations inside the basin can be observed for all scenarios (e.g. (B) and above (Wald et al 1999) Roermond earthquake). For wave fronts crossing the deepest parts of the basin focusing effects outside the basin margins are apparent Fig. 8 shows intensity maps for the earthquake scenarios. These for the Dürensimulation (C) resulting in amplified ground motion. maps clearly reveal amplification due to the 3-D basin effects, which Inside the basin channelling of energy is present (D). were apparent also in the snapshots. For the Dürensimulation high In order to further illustrate the basin effects, synthetic seismo- intensities outside the basin are visible and can be related to the grams are extracted on a profile along the entire basin. In Figs 4–6 shadow effect of waves passing through the deepest parts of the three component synthetic seismograms recorded at 30 virtual sta- basin (labelled B in Fig. 8a). However, this simulation reveals in tions along profile A–A (Fig. 2) are shown together with a profile general a strong correlation of intensity with basin depth (A). Chan- of basin depth. Recordings for each earthquake scenario are scaled nelling of wave energy identified for the Euskirchen simulation leads individually. To allow the comparison between the three scenarios to amplification in the corresponding regions (C in Fig. 8b). For individual PGV values are given in the top-right boxes. The influ- the Roermond earthquake strong correlation of intensity with basin ence of basin structure on both amplitude and shaking duration is depth is notable, especially on the small depressions southeast of notable on all components and for all scenarios. The most prominent the main basin (D in Fig. 8c). The most notable effect is a strong features are marked with labels (A–E): amplification of ground motion in a comparably shallow part of the The most obvious feature of the seismogram sections consists basin due to the refraction of wave energy at the deeper basins flanks of the higher amplitudes inside the basin (A) compared to those (E in Fig. 8c). The waves generating these amplitudes are also promi- recorded outside the basin (B) at the same distance. However, dif- nent in the snapshots (Fig. 3). Comparing this result to the Düren ferences in amplification as a function of source location are notable simulation, where these flanks worked as a boundary for the wave by comparing the individual scenarios. For the Roermond earth- energy (F), the sensitivity of basin response to source location with quake the region of strong amplification is much more restricted respect to prominent features of basin topography is demonstrated. (D) than for the other two examples. Prolonged shaking is present Generally, the simulations produce overestimated peak intensities for the Euskirchen scenario especially in a region between the two in the epicentral areas. deep depressions of the basin (C). The seismogram section for the For the 1992 Roermond earthquake a comparison of seismic in- Düren earthquake shows a single extremely large amplitude above tensities derived from synthetic PGV according to the formulae

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS 202 M. Ewald et al. Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

Figure 3. Snapshots of the simulated waveﬁeld (EW component) for three earthquake scenarios in the Lower Rhine Embayment (1756 D¨uren;1951 Euskirchen; 1992 Roermond, see text for details) at 6, 14, 22, 30 and 38 s simulation time. The covered region is the same as in Fig. 2. Amplitudes are scaled individually for each scenario. Complex wave front distortions due to the low-velocity structure of the basin are observable for all events. Note in particular the different extent of prolonged shaking due to reverberations inside the basin. Some regions discussed in text are indicated by capital letters.

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS Basin effects on ground motion in the LRE 203 Profile AÐA' Ð East/West 60 -1 50 1.38 m s

40 A B 30 E 20 Düren Time [s] 10

0 20 40 60 80 100 120 140 160 Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021 Profile Distance [km] 60 0.26 m s-1 50

40 30 C 20

10 Euskirchen 0

0 20 40 60 80 100 120 140 160 Profile Distance [km]

60 0.61 m s-1 50 40 D 30

Time [s]10 Time [s] Roermond 0

0 20 40 60 80 100 120 140 160 Profile Distance [km]

Figure 4. Velocity seismograms of the EW component along profile A–A (see Fig. 2) for the three simulated scenarios. The black line below zero indicates basin topography along the profile. Recordings for each earthquake scenario are scaled individually. Peak ground velocities along the profile are given for the individual scenarios in the top-right corner to allow comparison between the scenarios.

given above with the observed intensity distribution is presented 3-D basin effects can significantly increase durations of ground in Fig. 9. Area-wide synthetic seismic intensities are plotted in the motion. Several studies have found surface wave generation at basin same colour scale as in previous figures. Black lines indicate the edges particularly responsible for this phenomenon (e.g. Vidale & isoseismals derived from observations with seismic intensity levels Helmberger 1988; Olsen et al. 1995b). The basin effects on the dura- as indicated. These isoseismals are derived by Meidow & Ahorner tion of shaking are addressed with a simple approach: For each grid (1994) from a set of 2000 macroseismic reports at 600 different point at the surface the time is recorded at which each component locations. From this set of observations, also the isoseismal radii of the ground velocity exceeds a certain threshold value for the first shown in Fig. 4 were derived. The results are discussed in detail in time. The last time this value is again exceeded is also recorded. the following section. As threshold value a velocity of 4 cm s−1 is chosen (modified after

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS 204 M. Ewald et al.

60 -1 50 1.21 m s

20 Düren Time [s] 10

0 20 40 60 80 100 120 140 160 Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021 Profile Distance [km] 60 0.25 m s-1 50

10 Euskirchen 0

0 20 40 60 80 100 120 140 160 Profile Distance [km]

60 0.41 m s-1 50

Time [s]10 Time [s] Roermond 0

0 20 40 60 80 100 120 140 160 Profile Distance [km]

Figure 5. Same as Fig. 4 for the NS component velocity seismograms.

Olsen 2000), which corresponds to an intensity of IV–V accord- above strong gradients in basin topography, whereas another group ing to the relations given above. The length of the determined time of extended duration spots is due to the channelling effects notable window is called the shaking duration and mapped for each sce- also on snapshots and intensity maps (C). nario individually. In Fig. 10 isolines of shaking duration for the Although the number of simulated earthquakes is not sufficient for three simulated earthquakes are shown, colours indicating duration a representative seismic hazard evaluation, a first attempt is made to in seconds. Solid black lines depict the outline of the sedimentary combine results of the individual scenario simulations. Occurrence basin. frequencies for the three events are not considered and, therefore, the Except for the Dürensimulation the observation of enduring seismic intensity values from the simulations are equally weighted strong ground motion is restricted to the basin area. Besides this gen- and stacked. The resulting intensity distributions are called cumu- eral prolongation of shaking duration different spots of maximum lative intensity maps in the following. duration can be identified from the contour plots. Some of these Fig. 11 shows the evolution of this cumulative intensity map spots appear to be related to corners in basin topography wherein throughout the series of the simulated earthquakes. The series starts wave energy got trapped (A in Fig. 10). Spots labelled (B) appear with the intensity map from the simulation of the 1756 Düren

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS Basin effects on ground motion in the LRE 205 Profile AÐA' Ð Vertical 60 -1 50 0.87 m s

20 Düren Time [s] 10

0 20 40 60 80 100 120 140 160 Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021 Profile Distance [km] 60 0.24 m s-1 50

10 Euskirchen 0

0 20 40 60 80 100 120 140 160 Profile Distance [km]

60 0.22 m s-1 50

Time [s]10 Time [s] Roermond 0

0 20 40 60 80 100 120 140 160 Profile Distance [km]

Figure 6. Same as Fig. 4 for the vertical component ground velocity seismograms.

earthquake as shown before (Fig. 11a). The 1951 Euskirchen earth- probabilistic earthquake intensity isolines with an occurrence rate quake does not signiﬁcantly change the cumulative energy map. of 0.001 per year. Except for the region near Roermond, which is This is reasonable because epicentres are located relatively close dominated by the 1992 earthquake, the shape of the isolines roughly to each other and the magnitude of the 1951 event is distinctly matches. The values can only be compared qualitatively because in lower than the one from 1756 (Fig. 11b). After the 1992 Roermond the cumulative intensity map the occurrence rate is not taken into earthquake the northwestern part of the cumulative intensity map account. is dominated by this event. However, after this series of only three earthquakes, the source-related patterns of the cumulative intensity starts to change over into a more basin correlated shape (Fig. 11c). DISCUSSION The probabilistic seismic hazard model for the LRE presented by The simulated ground motions show distinct differences in ampli- Rosenhauer & Ahorner (1994) is compared to the cumulative in- ﬁcation on the three components of ground motion with varying tensity map (Fig. 11d). The black lines with numbers indicate the source location.

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS 206 M. Ewald et al.

station TGA (eastÐwest component) station TGA (northÐsouth component) 10 10 observed data observed data synthetic data synthetic data 8 8

6 6 ] ] -1 4 -1 4

2 2

0 0

Ð2 Ð2 Ground Velocity [cm s Ð4 Ground Velocity [cm s Ð4

Ð6 Ð6

Ð8 Ð8 Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

Ð10 Ð10 0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50 Time [s] Time [s]

station TGA (vertical component) 6

4 ] -1 2 [cm s

Ð2 Ground Velocity

Ð4

observed data synthetic data Ð6 0 5 10 15 20 25 30 35 40 45 50 Time [s]

Figure 7. Comparison of simulated (red lines) and observed (solid lines) seismograms at the station TGA (near Bergheim) for the ML5.9 1992 Roermond earthquake.

For the 1756 Dürenearthquake the maximum intensity from the 51 km, as it is derived from macroseismic maps (Meidow & Ahorner simulation is about one unit higher than the maximum intensity of 1994). This intensity level is well matched by the simulation result. VIII as estimated from historical records (Meidow 1994). In general, The observed epicentral intensities of I MM = VII–VIII are again the simulation results overestimate intensities in the near source area. significantly lower than the ones obtained by the simulations. For large parts of the basin, very large amplitudes and consequently The intensity maps for the 1992 Roermond earthquake shows high seismic intensities are produced above the basin edges. Two strong correlation of high intensities with the contours of the sed- explanations have to be considered: (1) Although the sedimentary imentary basin. The peak intensities from the simulations at the structure is modelled with a velocity gradient an overestimation of epicentre (IX) are higher than the reported ones, which are esti- the velocity contrast at the basin margin is still possible. Smoother mated after correction between VII and VIII (Meidow 1994). The horizontal velocity gradients or scattering may reduce focusing ef- mean isoseismal radii for I MM = VI (approximately 45 km) and fects above basin edges that are responsible for the large amplitudes. I MM = V (100 km) are demarked as black circles (Meidow 1994) and (2) The number of observations for the 1756 Dürenearthquake is correspond roughly with expected radii derived from the synthetic limited and the observations themselves may be uncertain. Infor- results. Meidow & Ahorner (1994) used a set of 2000 macroseismic mation on local disproportional high amplitudes might have been reports to derive isoseismals for the 1992 Roermond earthquake. In lost. However, the mean radii for I MM = VI, and I MM = V are better Fig. 9 these isoseismals (solid black lines with indicated intensity matched than the maximum intensities near the epicentre. Amplified levels) are compared with intensity maps derived from synthetic ground motions in a shadow zone north east of the basin are visible peak ground motion. both, in the snapshots and intensity maps. This effect is matched by Before discussing the considerable differences, it must be con- macroseismic observations from historical reports (Meidow 1994). sidered that seismic intensities compared herein are achieved with The intensity map of the 1951 Euskirchen earthquake shows a completely different approaches resulting in a number of limita- sharper distribution as it is the case for the other two scenarios tions, which are briefly discussed in the following. Whereas the due to lower seismic moment and focal depth. The black circle synthetic seismic intensities are derived from PGVs of the simulated (Fig. 4) demarks the mean isoseismal radius for intensity V with wavefield at each point, the observed isoseismals are derived from

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS Basin effects on ground motion in the LRE 207 Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

Figure 8. Modiﬁed Mercalli Intensity for the D¨uren,Euskirchen and Roermond earthquake calculated from simulated PGV values. Mean isoseismal radii from observations are marked as black circles (see text for details).

macroseismic reports, which intrinsically include effects like the local building conditions. In fact, the distribution of macroseismic reports reﬂects strongly the urban areas of the region, in particular south of the Roermond epicentre. As discussed above the numerical model for these simulations does not contain shallow structures particularly in the region close to the river Rhine. In consequence, individual site conditions in these areas are not included in the synthetics. Proper validation of the modelling results should be done with observations in the same frequency band as carried out by Ewald (2005). Due to the lack of observational data this cannot be done for the earthquakes investigated in this study. The macroseismic reports do not allow determining the period causing the reported shaking level. Maximum observed intensity VII and above is concentrated in a relatively small area of 92 km2 shifted eastward with respect to the instrumental epicentre (Meidow & Ahorner 1994). This area Figure 9. Comparison of synthetic and observed seismic intensities for is elongated in NW–SE direction and it coincides with the inner the 1992 Roermond, M 5.9 earthquake. Seismic intensities derived from ﬂank of the Peel boundary fault. The areas of highest intensity pro- synthetic peak ground velocities (see text for details) are colour-coded. Black duced by the simulations match this observation and the geometry lines indicate isoseismals derived from observations (Meidow 1994). supports the interpretation that these high levels of ground shaking

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS 208 M. Ewald et al. Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

Figure 10. Isolines of shaking duration for the simulated earthquakes. Colours indicate shaking duration in seconds. Solid black lines depict the shape of the sedimentary basin. Note the strong correlation between prolonged shaking and basin depth. are due to geometric amplification at the steep basin edges below From a comparison of the seismogram sections discussed above this area. The simulations overestimate absolute maximal values (Figs 4–6) it can be understood that basin-related amplification ef- of seismic intensity. It must be stated that the observed maximum fects show strong variability with source location. This is espe- levels of seismic intensity are considered as significantly too low, cially significant for the east–west component seismograms shown resulting in an estimation of hypocentre depth as deep as 26 km in Fig. 4. This variability has also a clear expression in above dis- from these observations compared to the instrumentally calculated cussed intensity maps derived from synthetic peak ground motion. depth of 17 km which was used for the simulation set-up (Meidow & The intensity maps in Fig. 8 show a stronger correlation between Ahorner 1994). Isoseismals for intensity VI are roughly matched by ground motion amplitude and basin depth for the Roermond event the simulation results except for the area with lower intensity north- than for the other simulated earthquakes. Seismogram sections of the east of the epicentre. The abrupt change in the shape of the related horizontal east–west component further demonstrate this behaviour. isoseismal in that area is constructed from only three macroseismic While the amplitudes of the Roermond earthquake show the high- reports and is not supported by the simulation results. Islands of est values above the deep basin depressions the ground motion of intensity VI south of the basins margins also refer to singular ob- the Dürenand Euskirchen earthquakes have maximum amplitudes servations and are not matched in particular by the synthetics. The above the steep basin edges. Scattering of wave energy is the dom- generally NW–SE trending shape of isoseismals for intensity V–VI inant effect for these recordings. Reverberations from basin edge are not recovered by the simulation results and the same accounts reflections dominate the duration of strong ground motion for the for the area of intensity VI south of the city of Bonn (see Figs 1 Euskirchen event. Because the amplitudes are generally lower for and 2). The observation of elevated ground motion in this area may this earthquake, the influences of the basin are more significant than be related to the presence of shallow soft sediments along the river for the other earthquakes. Reflected phases can be identified for both Rhine, which are not included in the velocity model used in this surface and body waves. For the Dürenearthquake with its greater study. magnitude and deeper hypocentre the basin influence is more visible

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS Basin effects on ground motion in the LRE 209 Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

Figure 11. Evolution of combined intensities throughout the sequence of the simulated three scenarios (a–c) and comparison with the current seismic hazard map for the study area (d). The black lines with numbers indicate the probabilistic earthquake intensity isolines with an occurrence rate of 0.001 yr−1 (Rosenhauer & Ahorner 1994).

in amplitudes than in duration. Due to the small epicentral distance, strong reverberations are present and single phases cannot easily be most of the effects in the northwestern part of the basin are invisible. identified from the NS component. A dominant onset on a single Strong influence can be seen at the southeastern depression of the station at receiver offset 60 km is notable. With respect to snapshots basin. and amplification maps, this can be explained by focusing effects The NS component (Fig. 5) of ground velocities in general display of the steep depression. Basin effects on the seismograms from the similar effects than the EW component. However, some interesting Roermond earthquake show strong correlation between amplitude differences appear. For the Dürenearthquake effects can be seen in and basin depth. The effects in shaking duration are not as dominant amplitude as well as in duration of shaking. Reflected phases from as for the other two earthquake scenarios. Note the difference in the basin margins can be identified. Although the duration of shaking amplitudes of shear and surface waves inside and outside of the is significantly prolonged, the wavefield is much more restricted in basin. time compared to the Euskirchen event. The largest amplitudes are Correlation of amplitudes with basin depth is not so obvious present 20–30 s after the first arrivals. For the Euskirchen event, for the vertical component ground motion as it is the case for the

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS 210 M. Ewald et al.

Table 4. Summary of source parameters for the 1992 Roermond earthquake from different studies. 16 Author Lat Lon Depth Strike Dip Rake M 0 [10 Nm] Braunmiller et al. (1994) 51.17 5.97 13 139±258±1 263±2 9.4±0.5 Braunmiller et al. (1994) 51.17 5.97 18 138±158±1 262±2 9.2±0.4 Ahorner (1994) 51.170 5.925 14.6 124 68 270 9.8 Paulssen et al. (1992) 51.17 5.97 20 124 70 270 — Dziewonski et al. (1993) 51.17 5.94 15 143 68 273 13.3 horizontal components (see Fig. 6). The main effects consist of wavefield complexity caused by the basin structure with strong de- longer shaking duration and generation of surface waves with high pendence on the location of the individual events. This emphasizes amplitudes. The vertical seismograms from the Dürenevent show the importance of accounting for both site and path effects in seismic longer shaking durations and less amplitude variations than the hor- hazard studies of basin regions. Furthermore, it can be concluded izontal component seismograms. that the use of station correction factors obtained by 1-D approaches For the Euskirchen event the change in frequency content men- may lead to errors in estimating earthquake source parameters, such Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021 tioned above for waves outside the basin is also visible (Fig. 4, as local magnitudes. middle panel D). For the Dürenevent a strong correlation of wave Although we found some agreement between synthetic and ob- amplitude with basin depth is notable. Scattering effects are not served waveforms for the one station where a direct comparison was as prominent as they were on the horizontal components. For the possible, strong motion records for further events from various sta- Roermond simulation, most of the effects are invisible due to the tions are necessary to check our simulations against observations. spherical correction used in the profile plots. This indicates them as On 2002 July 22 an M L 4.9 earthquake occurred within the study being weaker compared to horizontal component seismograms. area near the town of Alsdorf. Strong motion recordings of this event For the single station TGA where a direct comparison of synthetic on a recently installed seismic network in the Lower Rhine Embay- and observed seismograms is possible we achieve some significant ment were modelled with the same approach used in this study and similarities. First, the onsets of P and S waves are in good agree- allowed a detailed appraisal of the velocity model (Ewald 2005). ment. Secondly, the amplitude ratio of P and S waves is matched to Another limitation is the small number of three earthquake scenar- within 20 per cent. The maximum difference of about 50 per cent ios, which makes it possible only to give a flavour of the variability in absolute peak amplitudes is expected to be due to uncertainties of ground motion with different earthquake locations. Our results, on earthquake source parameters (see Table 4), as well as the fact especially the strong dependence of amplified ground motion on that our rheology model lacks the uppermost low-velocity layers. source location demonstrates the intrinsic 3-D nature of such ef- The fact, that peak amplitudes as well as some features such as fects and, therefore, confirm the need for 3-D simulations to assess the characteristics of late arriving energy on the vertical compo- shaking hazard in our study area. nent can be roughly reproduced demonstrate the potential of 3-D waveform modelling for hazard assessment in regions of moderate ACKNOWLEDGMENTS seismicity, where the lack of observations complicates shaking hazard estimates. Deriving the shaking duration from the observed and We thank the Leibniz Rechenzentrum in Munich for providing ac- synthetic seismograms according to the approach used in this study cess to their computational resources, especially the Hitachi SR8000 results in a value of 11.5 s for the observation and only 2.7 s for the supercomputer. We thank R. Wang (GFZ Potsdam) for providing his synthetics. However, it must be stated that this approach is designed reflectivity code that we used for testing the accuracy of our finite- in order to provide an area-wide rough estimate on this quantity at difference schemes. This work was funded partly by the MunichRE run time of the simulation. Inevitably, results are subject to large spa- Reinsurance Company and through the KONWIHR project (NBW). tial heterogeneities and are also depending in a non-linear fashion on the chosen threshold value. Visual comparison of the waveforms of the corresponding traces REFERENCES in Fig. 7 reveals that a small reduction of the threshold value would Ahorner, L., 1994. Fault-plane solutions and source parameters of the 1992 result in significantly better agreement between observations and Roermond, the Netherlands, mainshock and its stronger aftershocks from synthetics. From area-wide mapping of shaking duration shown in regional seismic data, Geologie en Mijnbouw (Special Issue: The Roer Fig. 10 it can be understood that for most locations close to the station Valley Graben, eds van Eck, T. & Davenport, C.A.), Vol. 73, pp. 199–214. TGA later arrivals contribute to the effective shaking duration and Boore, D.M., 1972. Finite Difference Methods for Seismic Wave Propaga- the estimates on this parameter are between 15 and 25 s. This is in tion in Heterogeneous Materials, in Methods in computational Physics, much better agreement with the observed value. Vol. 11, Academic, New York. Braunmiller, J., Dahm, T. & Bonjer, K.P., 1994. Source mechanism of the 1992 Roermond, the Netherlands, earthquake from inversion of regional surface waves (extended abstract), Geologie en Mijnbouw (Special Issue: CONCLUSION AND OUTLOOK The Roer Valley Graben, eds van Eck, T. & Davenport, C.A.), Vol. 73, In this study basin-related effects on waveform, PGV and shaking pp. 225–227. Brune, J.N., 1970. Tectonic stress and spectra of seismic shear waves from duration from earthquake ground motion in the Lower Rhine Em- earthquakes, J. geophys. Res., 76, 4997–5009. bayment are investigated. We employed an approach incorporating Cárdenas, M. & Chávez-Garcia, F.J., 2003. Regional path effects on seismic a fairly simple 3-D basin model and moment tensor finite source wave propagation in central Mexico, Bull. seism. Soc. Am., 93, 973–985. to be capable of simulating both recent and historical earthquake de Crook, T., 1994. Earthquake hazard for Roermond, the Netherlands, Ge- scenarios, the latter limiting the knowledge on source parameters. ologie en Mijnbouw (Special Issue: The Roer Valley Graben, eds van Eck, Even with the simple physical model used, we obtain a high level of T. & Davenport, C.A.), Vol. 73, pp. 425–429.

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS Basin effects on ground motion in the LRE 211

Dziewonski, A.M., Ekstr¨om, G. & Salganik, M., 1993. Centroid-moment Olsen, K.B., Archuleta, R.J. & Matarese, J.R., 1995b. Three-dimensional tensor solutions for April-June 1992, Phys. Earth planet. Int., 77, 151– simulation of a magnitude 7.75 earthquake on the San Andreas fault in 163. southern California, Science, 270, 1628–1632. Eisner, L. & Clayton, R.W., 2002. A full waveform test of the southern Olsen, K.B., Madariaga, R. & Archuleta, R.J., 1997. Three-dimensional dy- California velocity model by the reciprocity method, Pure appl. Geophys., namic simulation of the 1992 Landers earthquake, Science, 278, 834–838. 159, 1691–1706. Parolai, S., Bormann, P. & Milkereit, C., 2002. New relationships between Ewald, M., 2001. Numerical simulation of site effects with applications vs, thickness of sediments and resonance frequency calculated by the H/V to the Cologne basin, Diploma thesis, Institute of Geophysics, Ludwig- ratio of seismic noise for Cologne Area (Germany), Bull. seism. Soc. Am., Maximilians-University, Munich. 92, 2521–2527. Ewald, M., 2005. Numerical Simulations of Earthquake Scenarios in the Parolai, S., Picozzi, M., Richwalski, S.M. & Milkereit, B., 2005. Joint in- Lower Rhine Embayment Area, PHD thesis, Institute of Geophysics, version of phase velocity dispersion and H/V ratio curves from seismic Ludwig-Maximilians-University, Munich. noise recordings using a genetic algorithm, considering higher modes, Geluk, M.C., Duin, E.J.T., Dusar, M., Rijkers, R.H.B., van den Berg, M.W. Geophys. Res. Lett., 32(L01303), doi:10.1029/2004GL021115. & van Rooijen, P., 1994. Stratigraphy and tectonics of the Roer Valley Paulssen, H., Dost, B. & van Eck, T., 1992. The April 13, 1992 earthquake Graben, Geologie en Mijnbouw (Special Issue: The Roer Valley Graben, of Roermond (The Netherlands); ﬁrst interpretation of the NARS seismo-

eds van Eck, T. & Davenport, C.A.), Vol. 73, pp. 129–141. grams, Geol. Mijnbouw, 71, 91–98. Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021 Graves, R.W., 1996. Simulating seismic wave propagation in 3D elastic me- Pelzing, R., 1994. Source parameters of the 1992 Roermond earthquake, the dia using staggered grid-finite differences, Bull. seism. Soc. Am., 86, 1091– Netherlands, and some of its aftershocks recorded at stations of the Geo- 1106. logical Survey of Northrhine-Westphalia, Geologie en Mijnbouw (Special Graves, R.W.,1998. 3-D finite-difference modeling of the San Andreas Fault: Issue: The Roer Valley Graben, eds van Eck, T. & Davenport, C.A.), Vol. source parameterization and ground motion levels, Bull. seis. Soc. Amer., 73, pp. 215–223. 88, 881–897. Peyrat, S., Olsen, K.B. & Madariaga, R., 2001. Dynamic modeling of the Hinzen, K.-G., 2003. Stress field in the northern Rhine area, Central Europe, 1992 Landers earthquake, J. geophys. Res., 54, 468–488. from earthquake fault plane solutions, Tectonophysics, 377, 325–356. Pitarka, A., 1999. 3D elastic finite-difference modeling of seismic motion Hinzen, K.-G. & Oemisch, M., 2001. Location and magnitude from seismic using staggered grids with nonuniform spacing, Bull. seism. Soc. Am., 89, intensity data of recent and historic earthquakes in the northern Rhine 54–68. area, Central Europe, Bull. seism. Soc. Am., 91, 40–56. Pitarka, A. & Irikura, K., 1996a. Modeling 3D surface topography by a finite- Hinzen, K.-G., Scherbaum, F. & Weber, B., 2004. Study of the lateral res- difference method: Kobe-JMA station site, Japan, case study, Geophys. olution of H/V measurements across a normal fault in the Lower Rhine Res. Lett., 23, 2729–2732. Embayment, Germany, Journal of Earthquake Engineering, 8, 909–926. Pitarka, A. & Irikura, K., 1996b. Basin structure effects on long period strong Ibs-von Seht, M. & Wohlenberg, J., 1999. Microtremor measurements used motions in the San Fernando valley and the Los Angeles basin from the to map thickness of soft sediments, Bull. seism. Soc. Am., 89, 250–259. 1994 Northridge earthquake and aftershocks, Bull. seism. Soc. Am., 86, Igel, H., Mora, P.& Riollet, B., 1995. Anisotropic wave propagation through S126–S137. FD grids, Geophysics, 60, 1203–1216. Reamer, S.K. & Hinzen, K.-G., 2004. An earthquake catalog for the northern Joyner, W.B., 2000. Strong motion from surface waves in deep sedimentary Rhine area, Central Europe (1975–2002), Seism. Res. Lett., 75, 713–725. basins, Bull. seism. Soc. Am., 90, S95–S112. Robertsson, J.O.A., Blanch, J.O. & Symes, W.W., 1994. Viscoelastic finite- Kawase, H., 1996. The cause of damage belt in Kobe: “The basin-edge difference modeling, Geophysics, 59, 1444–1456. effect”, constructive interference of the direct S-waves with the basin Rosenhauer, W. & Ahorner, L., 1994. Seismic hazard assessment for the induced diffracted/Rayleigh waves, Seism. Res. Lett., 67, 25–34. Lower Rhine Embayment before and after the 1992 Roermond earthquake, Komatitsch, D., Qinya, L., Tromp, J., Süss,P., Stidham, C. & Shaw, J.H., Geologie en Mijnbouw (Special Issue: The Roer Valley Graben, eds van 2004. Simulations of ground motion in the Los Angeles basin based upon Eck, T. & Davenport, C.A.), Vol. 73, 415–424. the spectral-element method, Bull. seism. Soc. Am., 94, 187–206. Satoh, T., Kawase, H., Sato, T. & Pitarka, A., 2001. Three-dimensional finite- Magistrale, H., Graves, R. & Clayton, R., 1998. A standard three-dimensional difference waveform modeling of strong motions observed in the Sendai seismic velocity model for southern California: version 1, EOS, Trans. Am. basin, Japan, Bull. seism. Soc. Am., 91, 365–380. geophys. Un., 79, F605. Scherbaum, F., 1994. Modelling the Roermond earthquake of 1992 April Magistrale, H., Day, S., Clayton, R.W.& Graves, R., 2000. The SCEC South- 13 by stochastic simulation of its high-frequency strong ground motion, ern California reference three-dimensional seismic velocity model version Geophys. J. Int., 119, 31–43. 2, Bull. seism. Soc. Am., 90, S65–S76. Scherbaum, F., Hinzen, K.-G. & Ohrnberger, M., 2003. Determination of Meidow, H., 1994. Comparison of the macroseismic field of the 1992 Roer- shallow shear wave velocity profiles in the Cologne/Germany area using mond earthquake, the Netherlands, with those of large historical earth- ambient vibrations, Geophys. J. Int., 152, 597–612. quakes in the Lower Rhine Embayment and its vicinity, Geologie en Schwarzbach, M., 1951. Die Erdbeben des Rheinlandes, Kölner Geologische Mijnbouw (Special Issue: The Roer Valley Graben, eds van Eck, T. & Hefte, 1, 3–28. Davenport, C.A.), Vol. 73, 283–289. Scrivner, C.W. & Helmberger, D.V., 1994. Seismic Waveform Modeling in Meidow, H. & Ahorner, L., 1994. Macroseismic effects in Germany of the Los Angeles Basin, Bull. seism. Soc. Am., 84, 1310–1326. the 1992 Roermond earthquake and their interpretation, Stratigraphy and Sing, S.K. & Ordaz, M., 1993. On the origin of long coda observed in the tectonics of the Roer Valley Graben, Geologie en Mijnbouw (Special lake-bed strong-motion records of Mexico City, Bull. seism. Soc. Am., 83, Issue: The Roer Valley Graben, eds van Eck, T. & Davenport, C.A.), 1298–1306. Vol. 73, 271–279. Stidham, C., Antolik, M., Dreger, D., Larsen, S. & Romanovicz, B., 1999. Ohrnberger, M., Scherbaum, F., Krüger,F., Pelzing, R. & Reamer, S.K., Three-dimensional structure influence on the strong motion wavefield of 2004. How good are shear wave velocity models in the Lower Rhine Em- the 1989 Loma Prieta earthquake, Bull. seism. Soc. Am., 89, 1184–1202. bayment (NW-Germany) obtained from inversion of ambient vibrations?, Süss, M.P. & Shaw, J.H., 2003. P-wave seismic velocity structure derived Bolletino di Geofisica Teorica ed Applicata, 45(3), 215–232. from sonic logs and industry reflection data in the Los Angeles basin, Olsen, K.B., 2000. Site Amplification in the Los Angeles Basin from Three- California, J. geophys. Res., 108(B3), ESE13, 1–18. Dimensional Modeling of Ground Motion, Bull. seism. Soc. Am., 90, S77– Vidale, J.E. & Helmberger, D.V.,1988. Elastic finite-difference modeling of S94. the 1971 San Fernando, California, earthquake, Bull. seism. Soc. Am., 78, Olsen, K.B., Pechmann, J.C. & Schuster, G.T., 1995a. Simulation of 3-D 122–141. elastic wave propagation in the Salt Lake Basin, Bull. seism. Soc. Am., 85, Virieux, J., 1986. P-SV wave propagation in heterogeneous media: Velocity- 1688–1710. stress finite-difference method, Geophysics, 51, 889–901.

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS 212 M. Ewald et al.

Wald, D.J., Quitoriano, V., Heaton, T.H. & Kanamori, H., 1999. Relation- Basin for the Whittier-Narrows earthquake, Bull. seism. Soc. Am., 83, ships between peak ground acceleration, peak ground velocity and mod- 1325–1344. iﬁed Mercalli intensity In California, Earthquake Spectra, 15(3), 557– Ziegler, P.A., 1994. Cenozoic rift system of western and central Europe: an 564. overview, Geologie en Mijnbouw (Special Issue: The Roer Valley Graben, Yomogida, K. & Etgen, J.T., 1993. 3-D Wave propagation in the Los Angeles eds van Eck, T. & Davenport, C.A.), Vol. 73, 99–127. Downloaded from https://academic.oup.com/gji/article/166/1/197/632062 by guest on 24 September 2021

C 2006 The Authors, GJI, 166, 197–212 Journal compilation C 2006 RAS